The 'dumpability' of a task is fundamentally about the memory image of
the task - the concept comes from whether it can core dump or not - and
makes no sense when you don't have an associated mm.
And almost all users do in fact use it only for the case where the task
has a mm pointer.
But we have one odd special case: ptrace_may_access() uses 'dumpable' to
check various other things entirely independently of the MM (typically
explicitly using flags like PTRACE_MODE_READ_FSCREDS). Including for
threads that no longer have a VM (and maybe never did, like most kernel
threads).
It's not what this flag was designed for, but it is what it is.
The ptrace code does check that the uid/gid matches, so you do have to
be uid-0 to see kernel thread details, but this means that the
traditional "drop capabilities" model doesn't make any difference for
this all.
Make it all make a *bit* more sense by saying that if you don't have a
MM pointer, we'll use a cached "last dumpability" flag if the thread
ever had a MM (it will be zero for kernel threads since it is never
set), and require a proper CAP_SYS_PTRACE capability to override.
smb_inherit_dacl() walks the parent directory DACL loaded from the
security descriptor xattr. It verifies that each ACE contains the fixed
SID header before using it, but does not verify that the variable-length
SID described by sid.num_subauth is fully contained in the ACE.
A malformed inheritable ACE can advertise more subauthorities than are
present in the ACE. compare_sids() may then read past the ACE.
smb_set_ace() also clamps the copied destination SID, but used the
unchecked source SID count to compute the inherited ACE size. That could
advance the temporary inherited ACE buffer pointer and nt_size accounting
past the allocated buffer.
Fix this by validating the parent ACE SID count and SID length before
using the SID during inheritance. Compute the inherited ACE size from the
copied SID so the size matches the bounded destination SID. Reject the
inherited DACL if size accumulation would overflow smb_acl.size or the
security descriptor allocation size.
Fixes: e2f34481b24d ("cifsd: add server-side procedures for SMB3") Signed-off-by: Shota Zaizen <s@zaizen.me> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a field has been initialized, `init!`/`pin_init!` create a reference
or pinned reference to the field so it can be accessed later during the
initialization of other fields. However, the reference it created is
incorrectly `&'static` rather than just the scope of the initializer.
This is caused by `&mut (*#slot).#ident`, which actually allows arbitrary
lifetime, so this is effectively `'static`. Somewhat ironically, the safety
justification of creating the accessor is.. "SAFETY: TODO".
Fix it by adding `let_binding` method on `DropGuard` to shorten lifetime.
This results exactly what we want for these accessors. The safety and
invariant comments of `DropGuard` have been reworked; instead of reasoning
about what caller can do with the guard, express it in a way that the
ownership is transferred to the guard and `forget` takes it back, so the
unsafe operations within the `DropGuard` can be more easily justified.
Fixes: db96c5103ae6 ("add references to previously initialized fields") Signed-off-by: Gary Guo <gary@garyguo.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The CPU receives frames from the MAC through conventional DMA: the CPU
allocates buffers for the MAC, then the MAC fills them and returns
ownership to the CPU. For each hardware RX queue, the CPU and MAC
coordinate through a shared ring array of DMA descriptors: one
descriptor per DMA buffer. Each descriptor includes the buffer's
physical address and a status flag ("OWN") indicating which side owns
the buffer: OWN=0 for CPU, OWN=1 for MAC. The CPU is only allowed to set
the flag and the MAC is only allowed to clear it, and both must move
through the ring in sequence: thus the ring is used for both
"submissions" and "completions."
In the stmmac driver, stmmac_rx() bookmarks its position in the ring
with the `cur_rx` index. The main receive loop in that function checks
for rx_descs[cur_rx].own=0, gives the corresponding buffer to the
network stack (NULLing the pointer), and increments `cur_rx` modulo the
ring size. After the loop exits, stmmac_rx_refill(), which bookmarks its
position with `dirty_rx`, allocates fresh buffers and rearms the
descriptors (setting OWN=1). If it fails any allocation, it simply stops
early (leaving OWN=0) and will retry where it left off when next called.
This means descriptors have a three-stage lifecycle (terms my own):
- `empty` (OWN=1, buffer valid)
- `full` (OWN=0, buffer valid and populated)
- `dirty` (OWN=0, buffer NULL)
But because stmmac_rx() only checks OWN, it confuses `full`/`dirty`. In
the past (see 'Fixes:'), there was a bug where the loop could cycle
`cur_rx` all the way back to the first descriptor it dirtied, resulting
in a NULL dereference when mistaken for `full`. The aforementioned
commit resolved that *specific* failure by capping the loop's iteration
limit at `dma_rx_size - 1`, but this is only a partial fix: if the
previous stmmac_rx_refill() didn't complete, then there are leftover
`dirty` descriptors that the loop might encounter without needing to
cycle fully around. The current code therefore panics (see 'Closes:')
when stmmac_rx_refill() is memory-starved long enough for `cur_rx` to
catch up to `dirty_rx`.
Fix this by explicitly checking, before advancing `cur_rx`, if the next
entry is dirty; exit the loop if so. This prevents processing of the
final, used descriptor until stmmac_rx_refill() succeeds, but
fully prevents the `cur_rx == dirty_rx` ambiguity as the previous bugfix
intended: so remove the clamp as well. Since stmmac_rx_zc() is a
copy-paste-and-tweak of stmmac_rx() and the code structure is identical,
any fix to stmmac_rx() will also need a corresponding fix for
stmmac_rx_zc(). Therefore, apply the same check there.
In stmmac_rx() (not stmmac_rx_zc()), a related bug remains: after the
MAC sets OWN=0 on the final descriptor, it will be unable to send any
further DMA-complete IRQs until it's given more `empty` descriptors.
Currently, the driver simply *hopes* that the next stmmac_rx_refill()
succeeds, risking an indefinite stall of the receive process if not. But
this is not a regression, so it can be addressed in a future change.
Fixes: b6cb4541853c7 ("net: stmmac: avoid rx queue overrun") Closes: https://bugzilla.kernel.org/show_bug.cgi?id=221010 Cc: stable@vger.kernel.org Suggested-by: Russell King <linux@armlinux.org.uk> Signed-off-by: Sam Edwards <CFSworks@gmail.com> Link: https://patch.msgid.link/20260422044503.5349-1-CFSworks@gmail.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
STMMAC_GET_ENTRY() doesn't describe what this macro is doing - it is
incrementing the provided index for the circular array of descriptors.
Replace "GET" with "NEXT" as this better describes the action here.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Link: https://patch.msgid.link/E1w2vba-0000000DbWo-1oL5@rmk-PC.armlinux.org.uk Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 0bb05e6adfa9 ("net: stmmac: Prevent NULL deref when RX memory exhausted") Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use print_hex_dump_devel() for dumping sensitive HMAC key bytes in
hash_digest_key() to avoid leaking secrets at runtime when
CONFIG_DYNAMIC_DEBUG is enabled.
Add print_hex_dump_devel() as the hex dump equivalent of pr_devel(),
which emits output only when DEBUG is enabled, but keeps call sites
compiled otherwise.
Suggested-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev> Reviewed-by: John Ogness <john.ogness@linutronix.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Stable-dep-of: 177730a273b1 ("crypto: caam - guard HMAC key hex dumps in hash_digest_key") Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
vlan_dev_set_egress_priority() currently keeps cleared egress
priority mappings in the hash as tombstones. Repeated set/clear cycles
with distinct skb priorities therefore accumulate mapping nodes until
device teardown and leak memory.
Delete mappings when vlan_prio is cleared instead of keeping tombstones.
Now that the egress mapping lists are RCU protected, the node can be
unlinked safely and freed after a grace period.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Cc: stable@kernel.org Reported-by: Yifan Wu <yifanwucs@gmail.com> Reported-by: Juefei Pu <tomapufckgml@gmail.com> Reported-by: Xin Liu <bird@lzu.edu.cn> Co-developed-by: Yuan Tan <yuantan098@gmail.com> Signed-off-by: Yuan Tan <yuantan098@gmail.com> Signed-off-by: Longxuan Yu <ylong030@ucr.edu> Signed-off-by: Ren Wei <n05ec@lzu.edu.cn> Link: https://patch.msgid.link/ecfa6f6ce2467a42647ff4c5221238ae85b79a59.1776647968.git.yuantan098@gmail.com Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The TX fast path and reporting paths walk egress QoS mappings without
RTNL. Convert the mapping lists to RCU-protected pointers, use RCU
reader annotations in readers, and defer freeing mapping nodes with an
embedded rcu_head.
This prepares the egress QoS mapping code for safe removal of mapping
nodes in a follow-up change while preserving the current behavior.
hfsplus_fill_super() calls hfs_find_init() to initialize a search
structure, which acquires tree->tree_lock. If the subsequent call to
hfsplus_cat_build_key() fails, the function jumps to the out_put_root
error label without releasing the lock. The later cleanup path then
frees the tree data structure with the lock still held, triggering a
held lock freed warning.
Fix this by adding the missing hfs_find_exit(&fd) call before jumping
to the out_put_root error label. This ensures that tree->tree_lock is
properly released on the error path.
The bug was originally detected on v6.13-rc1 using an experimental
static analysis tool we are developing, and we have verified that the
issue persists in the latest mainline kernel. The tool is specifically
designed to detect memory management issues. It is currently under active
development and not yet publicly available.
We confirmed the bug by runtime testing under QEMU with x86_64 defconfig,
lockdep enabled, and CONFIG_HFSPLUS_FS=y. To trigger the error path, we
used GDB to dynamically shrink the max_unistr_len parameter to 1 before
hfsplus_asc2uni() is called. This forces hfsplus_asc2uni() to naturally
return -ENAMETOOLONG, which propagates to hfsplus_cat_build_key() and
exercises the faulty error path. The following warning was observed
during mount:
=========================
WARNING: held lock freed! 7.0.0-rc3-00016-gb4f0dd314b39 #4 Not tainted
-------------------------
mount/174 is freeing memory ffff888103f92000-ffff888103f92fff, with a lock still held there! ffff888103f920b0 (&tree->tree_lock){+.+.}-{4:4}, at: hfsplus_find_init+0x154/0x1e0
2 locks held by mount/174:
#0: ffff888103f960e0 (&type->s_umount_key#42/1){+.+.}-{4:4}, at: alloc_super.constprop.0+0x167/0xa40
#1: ffff888103f920b0 (&tree->tree_lock){+.+.}-{4:4}, at: hfsplus_find_init+0x154/0x1e0
Syzbot reported a KMSAN uninit-value issue in hfsplus_strcasecmp(). The
root cause is that hfs_brec_read() doesn't validate that the on-disk
record size matches the expected size for the record type being read.
When mounting a corrupted filesystem, hfs_brec_read() may read less data
than expected. For example, when reading a catalog thread record, the
debug output showed:
hfs_brec_read() only validates that entrylength is not greater than the
buffer size, but doesn't check if it's less than expected. It successfully
reads 26 bytes into a 520-byte structure and returns success, leaving 494
bytes uninitialized.
This uninitialized data in tmp.thread.nodeName then gets copied by
hfsplus_cat_build_key_uni() and used by hfsplus_strcasecmp(), triggering
the KMSAN warning when the uninitialized bytes are used as array indices
in case_fold().
Fix by introducing hfsplus_brec_read_cat() wrapper that:
1. Calls hfs_brec_read() to read the data
2. Validates the record size based on the type field:
- Fixed size for folder and file records
- Variable size for thread records (depends on string length)
3. Returns -EIO if size doesn't match expected
For thread records, check against HFSPLUS_MIN_THREAD_SZ before reading
nodeName.length to avoid reading uninitialized data at call sites that
don't zero-initialize the entry structure.
Also initialize the tmp variable in hfsplus_find_cat() as defensive
programming to ensure no uninitialized data even if validation is
bypassed.
QAT GEN6 hardware requires a minimum 3 us delay during the acceleration
engine reset sequence to ensure the hardware fully settles.
Without this delay, the firmware load may fail intermittently.
Add a delay after placing the AE into reset and before clearing the reset,
matching the hardware requirements and ensuring stable firmware loading.
Earlier generations remain unaffected.
The macros in qat_hal.c were using a mixture of tabs and spaces.
Update all macro indentation to use tabs consistently, matching the
predominant style.
Kingston eMMC IY2964 and IB2932 takes a fixed ~2 seconds for each secure
erase/trim operation regardless of size - that is, a single secure
erase/trim operation of 1MB takes the same time as 1GB. With default
calculated 3.5MB max discard size, secure erase 1GB requires ~300 separate
operations taking ~10 minutes total.
Add a card quirk, MMC_QUIRK_FIXED_SECURE_ERASE_TRIM_TIME, to set maximum
secure erase size for those devices. This allows 1GB secure erase to
complete in a single operation, reducing time from 10 minutes to just 2
seconds.
Signed-off-by: Luke Wang <ziniu.wang_1@nxp.com> Cc: stable@vger.kernel.org Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some eMMC vendors need to report manufacturing dates beyond 2025 but are
reluctant to update the EXT_CSD revision from 8 to 9. Changing the
Updating the EXT_CSD revision may involve additional testing or
qualification steps with customers. To ease this transition and avoid a
full re-qualification process, a workaround is needed. This
patch introduces a temporary quirk that re-purposes the year codes
corresponding to 2010, 2011, and 2012 to represent the years 2026, 2027,
and 2028, respectively. This solution is only valid for this three-year
period.
After 2028, vendors must update their firmware to set EXT_CSD_REV=9 to
continue reporting the correct manufacturing date in compliance with the
JEDEC standard.
The `MMC_QUIRK_BROKEN_MDT` is introduced and enabled for all Sandisk
devices to handle this behavior.
Signed-off-by: Avri Altman <avri.altman@sandisk.com> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Stable-dep-of: d6bf2e64dec8 ("mmc: core: Optimize time for secure erase/trim for some Kingston eMMCs") Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
JEDEC JESD84-B51B which was released in September 2025, increases the
manufacturing year limit for eMMC devices. The eMMC manufacturing year
is stored in a 4-bit field in the CID register. Originally, it covered
1997–2012. Later, with EXT_CSD_REV=8, it was extended up to 2025. Now,
with EXT_CSD_REV=9, the range is rolled over by another 16 years, up to
2038.
napi_build_skb() can return NULL on allocation failure. In
__octep_vf_oq_process_rx(), the result is used directly without a NULL
check in both the single-buffer and multi-fragment paths, leading to a
NULL pointer dereference.
Add NULL checks after both napi_build_skb() calls, properly advancing
descriptors and consuming remaining fragments on failure.
Fixes: 1cd3b407977c ("octeon_ep_vf: add Tx/Rx processing and interrupt support") Cc: stable@vger.kernel.org Signed-off-by: David Carlier <devnexen@gmail.com> Link: https://patch.msgid.link/20260409184009.930359-3-devnexen@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ inlined missing octep_vf_oq_next_idx() helper as read_idx++ with wraparound ] Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arena_vm_open() only bumps vml->mmap_count but never registers the
child VMA in arena->vma_list. The vml->vma always points at the
parent VMA, so after parent munmap the pointer dangles. If the child
then calls bpf_arena_free_pages(), zap_pages() reads the stale
vml->vma triggering use-after-free.
Fix this by preventing the arena VMA from being inherited across
fork with VM_DONTCOPY, and preventing VMA splits via the may_split
callback.
Also reject mremap with a .mremap callback returning -EINVAL. A
same-size mremap(MREMAP_FIXED) on the full arena VMA reaches
copy_vma() through the following path:
check_prep_vma() - returns 0 early: new_len == old_len
skips VM_DONTEXPAND check
prep_move_vma() - vm_start == old_addr and
vm_end == old_addr + old_len
so may_split is never called
move_vma()
copy_vma_and_data()
copy_vma()
vm_area_dup() - copies vm_private_data (vml pointer)
vm_ops->open() - bumps vml->mmap_count
vm_ops->mremap() - returns -EINVAL, rollback unmaps new VMA
The refcount ensures the rollback's arena_vm_close does not free
the vml shared with the original VMA.
Reported-by: Weiming Shi <bestswngs@gmail.com> Reported-by: Xiang Mei <xmei5@asu.edu> Fixes: 317460317a02 ("bpf: Introduce bpf_arena.") Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com> Link: https://lore.kernel.org/r/20260413194245.21449-1-alexei.starovoitov@gmail.com Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When firmware enables 64-bit PCI host bridge support, some root bridges
already provide valid 64-bit mem resource windows through ACPI.
In this case, the LoongArch-specific mem resource high-bits fixup in
acpi_prepare_root_resources() should not be applied unconditionally.
Otherwise, the kernel may override the native resource layout derived
from firmware, and later BAR assignment can fail to place device BARs
into the intended 64-bit address space correctly.
Add a per-root-bridge ACPI flag, PCIH, and evaluate it from the current
root bridge device scope. When PCIH is set, skip the mem resource high-
bits fixup path and let the kernel use the firmware-provided resource
description directly. When PCIH is absent or cleared, keep the existing
behavior and continue filling the high address bits from the host bridge
address.
This makes the behavior per-root-bridge configurable and avoids breaking
valid 64-bit BAR space allocation on bridges whose 64-bit windows have
already been fully described by firmware.
kvm_flush_pte() is the only caller that directly assigns *pte instead
of using the kvm_set_pte() wrapper. Use the wrapper for consistency with
the rest of the file.
No functional change intended.
Cc: stable@vger.kernel.org Reviewed-by: Bibo Mao <maobibo@loongson.cn> Signed-off-by: Tao Cui <cuitao@kylinos.cn> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When timer interrupt arrives in guest kernel, guest kernel clears the
timer interrupt and program timer with the next incoming event.
During this stage, timer tick is -1 and timer interrupt status is
disabled in ESTAT register. KVM hypervisor need write zero with timer
tick register and wait timer interrupt injection from HW side, and
then clear timer interrupt.
So there is 2 cycle delay in KVM hypervisor to emulate such scenery,
and the delay is unnecessary if there is no need to clear the timer
interrupt.
Here move 2 cycle delay into timer clear scenery and add timer ESTAT
checking after delay, and set max timer expire value if timer interrupt
does not arrive still.
With passthrough HW timer, timer interrupt is injected by HW. When
inject emulated CPU interrupt by software such SIP0/SIP1/IPI, HW timer
interrupt may be lost.
Here check whether there is timer tick value inversion before and after
injecting emulated CPU interrupt by software, timer enabling by reading
timer cfg register is skipped. If the timer tick value is detected with
changing, then timer should be enabled. And inject a timer interrupt by
software if there is.
Insert the appropriate UNWIND hint into the kvm_exc_entry assembly
function to guide the generation of correct ORC table entries, thereby
solving the timeout problem ("unreliable stack") while loading the
livepatch-sample module on a physical machine running virtual machines
with multiple vcpus.
The switch case in loongson_gpu_fixup_dma_hang() may not DC2 or DC3, and
readl(crtc_reg) will access with random address, because the "device" is
from "base+PCI_DEVICE_ID", "base" is from "pdev->devfn+1". This is wrong
when my platform inserts a discrete GPU:
1. If a check fails after hyp_pin_shared_mem() succeeds, the cleanup
path jumps to 'unlock' without calling unpin_host_vcpu() or
unpin_host_sve_state(), permanently leaking pin references on the
host vCPU and SVE state pages.
Extract a register_hyp_vcpu() helper that performs the checks and
the store. When register_hyp_vcpu() returns an error, call
unpin_host_vcpu() and unpin_host_sve_state() inline before falling
through to the existing 'unlock' label.
2. register_hyp_vcpu() publishes the new vCPU pointer into
'hyp_vm->vcpus[]' with a bare store, allowing a concurrent caller
of pkvm_load_hyp_vcpu() to observe a partially initialised vCPU
object.
Ensure the store uses smp_store_release() and the load uses
smp_load_acquire(). While 'vm_table_lock' currently serialises the
store and the load, these barriers ensure the reader sees the fully
initialised 'hyp_vcpu' object even if there were a lockless path or
if the lock's own ordering guarantees were insufficient for nested
object initialization.
Fixes: 49af6ddb8e5c ("KVM: arm64: Add infrastructure to create and track pKVM instances at EL2") Reported-by: Ben Simner <ben.simner@cl.cam.ac.uk> Co-developed-by: Will Deacon <willdeacon@google.com> Signed-off-by: Will Deacon <willdeacon@google.com> Signed-off-by: Fuad Tabba <tabba@google.com> Link: https://patch.msgid.link/20260424084908.370776-6-tabba@google.com Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
FEAT_Debugv8p9 is incorrectly defined against ID_AA64DFR0_EL1.PMUVer
instead of ID_AA64DFR0_EL1.DebugVer. All three consumers of the macro
gate features that are architecturally tied to FEAT_Debugv8p9
(DebugVer = 0b1011, DDI0487 M.b A2.2.10):
- HDFGRTR2_EL2.nMDSELR_EL1, HDFGWTR2_EL2.nMDSELR_EL1: MDSELR_EL1
is present only when FEAT_Debugv8p9 is implemented (D24.3.21).
- MDCR_EL2.EBWE: the Extended Breakpoint and Watchpoint Enable bit
is RES0 unless FEAT_Debugv8p9 is implemented (D24.3.17).
Neither register has any dependency on PMUVer.
FEAT_Debugv8p9 and FEAT_PMUv3p9 are independent. Per DDI0487 M.b
A2.2.10, FEAT_Debugv8p9 is unconditionally mandatory from Armv8.9,
whereas FEAT_PMUv3p9 is mandatory only when FEAT_PMUv3 is implemented.
An Armv8.9 CPU without a PMU has DebugVer = 0b1011 but PMUVer = 0b0000,
so the wrong field check would cause KVM to incorrectly treat EBWE and
MDSELR_EL1 as RES0 on such hardware.
FEAT_SPE_FnE is architecturally detected via PMSIDR_EL1.FnE [6], not
ID_AA64DFR0_EL1.PMSVer. The FEAT_X macro form (register, field, value)
cannot encode a PMSIDR_EL1-based feature, so FEAT_SPE_FnE was defined
identically to FEAT_SPEv1p2 (ID_AA64DFR0_EL1, PMSVer, V1P2), producing
a duplicate that used PMSVer >= V1P2 as a proxy.
Replace the macro with feat_spe_fne(), following the same pattern as
the sibling feat_spe_fds(): guard on FEAT_SPEv1p2 and read
PMSIDR_EL1.FnE [6] directly. Wire the two NEEDS_FEAT consumers to use
the new function.
fix_host_ownership() walks the hypervisor's stage-1 page-table to
adjust the host's stage-2 accordingly. Any such adjustment that
requires cache maintenance operations depends on the per-CPU hyp
fixmap being present. However, fix_host_ownership() is currently
called before fix_hyp_pgtable_refcnt() and hyp_create_fixmap(), so
the fixmap does not yet exist when it runs.
This is benign today because the host stage-2 starts empty and no
CMOs are needed, but it becomes a latent crash as soon as
fix_host_ownership() is extended to operate on a non-empty
page-table.
Reorder the calls so that fix_hyp_pgtable_refcnt() and
hyp_create_fixmap() complete before fix_host_ownership() is invoked.
Fixes: 0d16d12eb26e ("KVM: arm64: Fix-up hyp stage-1 refcounts for all pages mapped at EL2") Signed-off-by: Quentin Perret <qperret@google.com> Signed-off-by: Fuad Tabba <tabba@google.com> Link: https://patch.msgid.link/20260424084908.370776-7-tabba@google.com Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The uaccess write handlers for GICD_IIDR in both GICv2 and GICv3
extract the revision field from 'reg' (the current IIDR value read back
from the emulated distributor) instead of 'val' (the value userspace is
trying to write). This means userspace can never actually change the
implementation revision — the extracted value is always the current one.
Fix the FIELD_GET to use 'val' so that userspace can select a different
revision for migration compatibility.
Fixes: 49a1a2c70a7f ("KVM: arm64: vgic-v3: Advertise GICR_CTLR.{IR, CES} as a new GICD_IIDR revision") Signed-off-by: David Woodhouse <dwmw@amazon.co.uk> Link: https://patch.msgid.link/20260407210949.2076251-2-dwmw2@infradead.org Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It appears that there is nothing in the wake-up path that
evaluates whether the in-kernel interrupts are pending unless
we have a vgic.
This means that the userspace irqchip support has been broken for
about four years, and nobody noticed. It was also broken before
as we wouldn't wake-up on a PMU interrupt, but hey, who cares...
It is probably time to remove the feature altogether, because it
was a terrible idea 10 years ago, and it still is.
During FGGC node block migration, fsck may incorrectly treat the
migrated node block as fsync-written data.
The reproduction scenario:
root@vm:/mnt/f2fs# seq 1 2048 | xargs -n 1 ./test_sync // write inline inode and sync
root@vm:/mnt/f2fs# rm -f 1
root@vm:/mnt/f2fs# sync
root@vm:/mnt/f2fs# f2fs_io gc_range // move data block in sync mode and not write CP
SPO, "fsck --dry-run" find inode has already checkpointed but still
with DENT_BIT_SHIFT set
The root cause is that GC does not clear the dentry mark and fsync mark
during node block migration, leading fsck to misinterpret them as
user-issued fsync writes.
In BGGC mode, node block migration is handled by f2fs_sync_node_pages(),
which guarantees the dentry and fsync marks are cleared before writing.
This patch move the set/clear of the fsync|dentry marks into
__write_node_folio to make the logic clearer, and ensures the
fsync|dentry mark is cleared in FGGC.
Cc: stable@kernel.org Fixes: da011cc0da8c ("f2fs: move node pages only in victim section during GC") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When f2fs_fiemap() is called with `fileinfo->fi_flags` containing the
FIEMAP_FLAG_SYNC flag, it attempts to write data to disk before
retrieving file mappings via filemap_write_and_wait(). However, there is
an issue where the file does not get mapped as expected. The following
scenario can occur:
The root cause of this issue is that f2fs_write_single_data_page() only
calls f2fs_write_inline_data() to copy data from the data folio to the
inode folio, and it clears the dirty flag on the data folio. However, it
does not mark the data folio as writeback. When
__filemap_fdatawait_range() checks for folios with the writeback flag,
it returns early, causing f2fs_fiemap() to report that the file has no
mapping.
To fix this issue, the solution is to call
f2fs_write_single_node_folio() in f2fs_inline_data_fiemap() when
getting fiemap with FIEMAP_FLAG_SYNC flags. This patch ensures that the
inode folio is written back and the writeback process completes before
proceeding.
Cc: stable@kernel.org Fixes: 9ffe0fb5f3bb ("f2fs: handle inline data operations") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
f2fs_destroy_extent_node() does not set FI_NO_EXTENT before clearing
extent nodes. When called from f2fs_drop_inode() with I_SYNC set,
concurrent kworker writeback can insert new extent nodes into the same
extent tree, racing with the destroy and triggering f2fs_bug_on() in
__destroy_extent_node(). The scenario is as follows:
drop inode writeback
- iput
- f2fs_drop_inode // I_SYNC set
- f2fs_destroy_extent_node
- __destroy_extent_node
- while (node_cnt) {
write_lock(&et->lock)
__free_extent_tree
write_unlock(&et->lock)
- __writeback_single_inode
- f2fs_outplace_write_data
- f2fs_update_read_extent_cache
- __update_extent_tree_range
// FI_NO_EXTENT not set,
// insert new extent node
} // node_cnt == 0, exit while
- f2fs_bug_on(node_cnt) // node_cnt > 0
Additionally, __update_extent_tree_range() only checks FI_NO_EXTENT for
EX_READ type, leaving EX_BLOCK_AGE updates completely unprotected.
This patch set FI_NO_EXTENT under et->lock in __destroy_extent_node(),
consistent with other callers (__update_extent_tree_range and
__drop_extent_tree) and check FI_NO_EXTENT for both EX_READ and
EX_BLOCK_AGE tree.
Fixes: 3fc5d5a182f6 ("f2fs: fix to shrink read extent node in batches") Cc: stable@vger.kernel.org Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When f2fs_map_blocks()->f2fs_map_blocks_cached() hits the read extent
cache, map->m_multidev_dio is not updated, which leads to incorrect
multidevice information being reported by trace_f2fs_map_blocks().
This patch updates map->m_multidev_dio in f2fs_map_blocks_cached() when
the read extent cache is hit.
Cc: stable@kernel.org Fixes: 0094e98bd147 ("f2fs: factor a f2fs_map_blocks_cached helper") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When `fileinfo->fi_flags` does not have the `FIEMAP_FLAG_SYNC` bit set
and inline data has not been persisted yet, the physical address of the
extent is calculated incorrectly for unwritten inline inodes.
This patch fixes the issue by checking if the inode's address is valid.
If the inline inode is unwritten, set the physical address to 0 and
mark the extent with `FIEMAP_EXTENT_UNKNOWN | FIEMAP_EXTENT_DELALLOC`
flags.
Cc: stable@kernel.org Fixes: 67f8cf3cee6f ("f2fs: support fiemap for inline_data") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
f2fs_need_dentry_mark() reads nat_entry flags without mutual exclusion
with the checkpoint path, which can result in an incorrect inode block
marking state. The scenario is as follows:
create & write & fsync 'file A' write checkpoint
- f2fs_do_sync_file // inline inode
- f2fs_write_inode // inode folio is dirty
- f2fs_write_checkpoint
- f2fs_flush_merged_writes
- f2fs_sync_node_pages
- f2fs_fsync_node_pages // no dirty node
- f2fs_need_inode_block_update // return true
- f2fs_fsync_node_pages // inode dirtied
- f2fs_need_dentry_mark //return true
- f2fs_flush_nat_entries
- f2fs_write_checkpoint end
- __write_node_folio // inode with DENT_BIT_SHIFT set
SPO, "fsck --dry-run" find inode has already checkpointed but still
with DENT_BIT_SHIFT set
The state observed by f2fs_need_dentry_mark() can differ from the state
observed in __write_node_folio() after acquiring sbi->node_write. The
root cause is that the semantics of IS_CHECKPOINTED and
HAS_FSYNCED_INODE are only guaranteed after the checkpoint write has
fully completed.
This patch moves set_dentry_mark() into __write_node_folio() and
protects it with the sbi->node_write lock.
Cc: stable@kernel.org Fixes: 88bd02c9472a ("f2fs: fix conditions to remain recovery information in f2fs_sync_file") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
f2fs_fiemap() calls f2fs_map_blocks() to obtain the block mapping a
file, and then merges contiguous mappings into extents. If the mapping
is found in the read extent cache, node blocks do not need to be read.
However, in the following scenario, a contiguous extent can be split
into two extents:
Although the physical addresses of the ranges 0～2MB and 2M～4MB are
contiguous, the mapping for the 2M～4MB range is not present in memory.
When the physical addresses for the 0～2MB range are updated, no merge
happens because the adjacent mapping is missing from the in-memory
cache. As a result, fiemap reports two separate extents instead of a
single contiguous one.
The root cause is that the read extent cache does not guarantee that all
blocks of an extent are present in memory. Therefore, when the extent
length returned by f2fs_map_blocks_cached() is smaller than maxblocks,
the remaining mappings are retrieved via f2fs_get_dnode_of_data() to
ensure correct fiemap extent boundary handling.
Cc: stable@kernel.org Fixes: cd8fc5226bef ("f2fs: remove the create argument to f2fs_map_blocks") Signed-off-by: Yongpeng Yang <yangyongpeng@xiaomi.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
a) TCMU device removal context:
- call del_gendisk() to get q->q_usage_counter
- call start_flush_work() to get work_completion of wb->dwork
b) f2fs writeback context:
- in wb_workfn(), which holds work_completion of wb->dwork
- call f2fs_balance_fs() to get sbi->gc_lock
c) f2fs vfs_write context:
- call f2fs_gc() to get sbi->gc_lock
- call f2fs_write_checkpoint() to get sbi->cp_global_sem
d) f2fs mount context:
- call recover_fsync_data() to get sbi->cp_global_sem
- call f2fs_check_and_fix_write_pointer() to call blkdev_report_zones()
that goes down to blk_mq_alloc_request and get q->q_usage_counter
Original callstack is in Closes tag.
However, I think this is a false alarm due to before mount returns
successfully (context d), we can not access file therein via vfs_write
(context c).
Let's introduce per-sb cp_global_sem_key, and assign the key for
cp_global_sem, so that lockdep can recognize cp_global_sem from
different super block correctly.
A lot of work are done by Shin'ichiro Kawasaki, thanks a lot for
the work.
f2fs_update_inode() reads inode->i_blocks without holding i_lock to
serialize it to the on-disk inode, while concurrent truncate or
allocation paths may modify i_blocks under i_lock. Since blkcnt_t is
u64, this risks torn reads on 32-bit architectures.
Following the approach in ext4_inode_blocks_set(), add READ_ONCE() to prevent
potential compiler-induced tearing.
When an ADD_ADDR needs to be retransmitted and another one has already
been prepared -- e.g. multiple ADD_ADDRs have been sent in a row and
need to be retransmitted later -- this additional retransmission will
need to wait.
In this case, the timer was reset to TCP_RTO_MAX / 8, which is ~15
seconds. This delay is unnecessary long: it should just be rescheduled
at the next opportunity, e.g. after the retransmission timeout.
Without this modification, some issues can be seen from time to time in
the selftests when multiple ADD_ADDRs are sent, and the host takes time
to process them, e.g. the "signal addresses, ADD_ADDR timeout" MPTCP
Join selftest, especially with a debug kernel config.
Note that on older kernels, 'timeout' is not available. It should be
enough to replace it by one second (HZ).
Fixes: 00cfd77b9063 ("mptcp: retransmit ADD_ADDR when timeout") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-6-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When an ADD_ADDR is retransmitted, the sk is held in sk_reset_timer(),
and released at the end.
If at that moment, it was the last reference being held, the sk would
not be freed. sock_put() should then be called instead of __sock_put().
But that's not enough: if it is the last reference, sock_put() will call
sk_free(), which will end up calling sk_stop_timer_sync() on the same
timer, and waiting indefinitely to finish. So it is needed to mark that
the timer is done at the end of the timer handler when it has not been
rescheduled, not to call sk_stop_timer_sync() on "itself".
Fixes: 00cfd77b9063 ("mptcp: retransmit ADD_ADDR when timeout") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-5-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When an ADD_ADDR is retransmitted, the sk is held in sk_reset_timer().
It should then be released in all cases at the end.
Some (unlikely) checks were returning directly instead of calling
sock_put() to decrease the refcount. Jump to a new 'exit' label to call
__sock_put() (which will become sock_put() in the next commit) to fix
this potential leak.
While at it, drop the '!msk' check which cannot happen because it is
never reset, and explicitly mark the remaining one as "unlikely".
Fixes: 00cfd77b9063 ("mptcp: retransmit ADD_ADDR when timeout") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-4-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This mptcp_pm_add_timer() helper is executed as a timer callback in
softirq context. To avoid any data races, the socket lock needs to be
held with bh_lock_sock().
If the socket is in use, retry again soon after, similar to what is done
with the keepalive timer.
Fixes: 00cfd77b9063 ("mptcp: retransmit ADD_ADDR when timeout") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-3-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ADD_ADDR can be sent for the ID 0, which corresponds to the local
address and port linked to the initial subflow.
Indeed, this address could be removed, and re-added later on, e.g. what
is done in the "delete re-add signal" MPTCP Join selftests. So no reason
to ignore it.
Fixes: 00cfd77b9063 ("mptcp: retransmit ADD_ADDR when timeout") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-2-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When adding the ADD_ADDR to the list, the address including the IP, port
and ID are copied. On the other hand, when the endpoint corresponds to
the one from the initial subflow, the ID is set to 0, as specified by
the MPTCP protocol.
The issue is that the ID was reset after having copied the ID in the
ADD_ADDR entry. So the retransmission was done, but using a different ID
than the initial one.
Fixes: 8b8ed1b429f8 ("mptcp: pm: reuse ID 0 after delete and re-add") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-1-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This variable counts how many MPTCP endpoints have a 'fullmesh' flag
set. After having flushed all MPTCP endpoints, it is then needed to
reset this counter.
Without this reset, this counter exposed to the userspace is wrong, but
also non-fullmesh endpoints added after the flush will not be taken into
account to create subflows in reaction to ADD_ADDRs.
mptcp_setsockopt_all_sf() was missing a call to sockopt_seq_inc(). This
is required not to cause missing synchronization for newer subflows
created later on.
This helper is called each time a socket option is set on subflows, and
future ones will need to inherit this option after their creation.
Both mptcp_setsockopt_sol_socket_tstamp() and
mptcp_setsockopt_sol_socket_timestamping() iterate over subflows,
acquire the subflow socket lock, but then erroneously pass the MPTCP
msk socket to sock_set_timestamp() / sock_set_timestamping() instead
of the subflow ssk. As a result, the timestamp flags are set on the
wrong socket and have no effect on the actual subflows.
Pass ssk instead of sk to both helpers.
Fixes: 9061f24bf82e ("mptcp: sockopt: propagate timestamp request to subflows") Cc: stable@vger.kernel.org Signed-off-by: Gang Yan <yangang@kylinos.cn> Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260427-net-mptcp-misc-fixes-7-1-rc2-v1-1-7432b7f279fa@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When HMAC validation fails on a received ACK + MP_JOIN in
subflow_syn_recv_sock(), the subflow is reset with reason
MPTCP_RST_EPROHIBIT ("Administratively prohibited"). This is
incorrect: HMAC validation failure is an MPTCP protocol-level
error, not an administrative policy denial.
The mirror site on the client, in subflow_finish_connect(), already
uses MPTCP_RST_EMPTCP ("MPTCP-specific error") for the same kind of
HMAC failure on the SYN/ACK + MP_JOIN. Use the same reason on the
server side for symmetry and accuracy.
In subflow_finish_connect(), HMAC validation of the server's HMAC
in SYN/ACK + MP_JOIN increments MPTCP_MIB_JOINACKMAC ("HMAC was
wrong on ACK + MP_JOIN") on failure. The function processes the
SYN/ACK, not the ACK; the matching MPTCP_MIB_JOINSYNACKMAC counter
("HMAC was wrong on SYN/ACK + MP_JOIN") exists but is not
incremented anywhere in the tree.
The mirror site on the server, subflow_syn_recv_sock(), already
uses JOINACKMAC correctly for ACK HMAC failure. Use JOINSYNACKMAC
at the SYN/ACK validation site so each counter reflects the packet
whose HMAC actually failed.
The SO_LINGER socket option has been supported for a while with MPTCP
sockets [1], but it didn't cause the equivalent of a TCP reset as
expected when enabled and its time was set to 0. This was causing some
behavioural differences with TCP where some connections were not
promptly stopped as expected.
To fix that, an extra condition is checked at close() time before
sending an MP_FASTCLOSE, the MPTCP equivalent of a TCP reset.
Note that backporting up to [1] will be difficult as more changes are
needed to be able to send MP_FASTCLOSE. It seems better to stop at [2],
which was supposed to already imitate TCP.
Validated with MPTCP packetdrill tests [3].
Fixes: 268b12387460 ("mptcp: setsockopt: support SO_LINGER") [1] Fixes: d21f83485518 ("mptcp: use fastclose on more edge scenarios") [2] Cc: stable@vger.kernel.org Reported-by: Lance Tuller <lance@lance0.com> Closes: https://github.com/lance0/xfr/pull/67 Link: https://github.com/multipath-tcp/packetdrill/pull/196 Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260427-net-mptcp-misc-fixes-7-1-rc2-v1-3-7432b7f279fa@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When pm_netlink.sh is executed with '-i', 'ip mptcp' is used instead of
'pm_nl_ctl'. IPRoute2 doesn't support the 'unknown' flag, which has only
been added to 'pm_nl_ctl' for this specific check: to ensure that the
kernel ignores such unsupported flag.
No reason to add this flag to 'ip mptcp'. Then, this check should be
skipped when 'ip mptcp' is used.
Fixes: 0cef6fcac24d ("selftests: mptcp: ip_mptcp option for more scripts") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau <martineau@kernel.org> Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20260505-net-mptcp-pm-fixes-7-1-rc3-v1-11-fca8091060a4@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Using '${?}' inside the if-statement to check the returned value from
the command that was evaluated as part of the if-statement is not
correct: here, '${?}' will be linked to the previous instruction, not
the one that is expected here (${cmd}).
Instead, simply mark the error, except if an error is expected. If
that's the case, 1 can be passed as the 4th argument of this helper.
Three checks from pm_netlink.sh expect an error.
While at it, improve the error message when the command unexpectedly
fails or succeeds.
Note that we could expect a specific returned value, but the checks
currently expecting an error can be used with 'ip mptcp' or 'pm_nl_ctl',
and these two tools don't return the same error code.
dispatch_enqueue()'s FIFO-tail path used list_empty(&dsq->list) to decide
whether to set dsq->first_task on enqueue. dsq->list can contain parked BPF
iterator cursors (SCX_DSQ_LNODE_ITER_CURSOR), so list_empty() is not a
reliable "no real task" check. If the last real task is unlinked while a
cursor is parked, first_task becomes NULL; the next FIFO-tail enqueue then
sees list_empty() == false and skips the first_task update, leaving
scx_bpf_dsq_peek() returning NULL for a non-empty DSQ.
Test dsq->first_task directly, which already tracks only real tasks and is
maintained under dsq->lock.
Fixes: 44f5c8ec5b9a ("sched_ext: Add lockless peek operation for DSQs") Cc: stable@vger.kernel.org # v6.19+ Reported-by: Chris Mason <clm@meta.com> Signed-off-by: Tejun Heo <tj@kernel.org> Reviewed-by: Andrea Righi <arighi@nvidia.com> Cc: Ryan Newton <newton@meta.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
scx_select_cpu_dfl() narrows @allowed to @cpus_allowed & @p->cpus_ptr
when the BPF caller supplies a @cpus_allowed that differs from
@p->cpus_ptr and @p doesn't have full affinity. However,
@is_prev_allowed was computed against the original (wider)
@cpus_allowed, so the prev_cpu fast paths could pick a @prev_cpu that
is in @cpus_allowed but not in @p->cpus_ptr, violating the intended
invariant that the returned CPU is always usable by @p. The kernel
masks this via the SCX_EV_SELECT_CPU_FALLBACK fallback, but the
behavior contradicts the documented contract.
Move the @is_prev_allowed evaluation past the narrowing block so it
tests against the final @allowed mask.
Fixes: ee9a4e92799d ("sched_ext: idle: Properly handle invalid prev_cpu during idle selection") Cc: stable@vger.kernel.org # v6.16+ Assisted-by: Claude <noreply@anthropic.com> Signed-off-by: David Carlier <devnexen@gmail.com> Reviewed-by: Andrea Righi <arighi@nvidia.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Even after applying commit 7244491dab34 ("RDMA/rxe: Validate pad and ICRC
before payload_size() in rxe_rcv"), a single unauthenticated UDP packet
can still trigger panic. That patch handled payload_size() underflow only
for valid opcodes with short packets, not for packets carrying an unknown
opcode. The unknown-opcode OOB read described below predates that commit
and reaches back to the initial Soft RoCE driver.
where header_size(pkt) expands to rxe_opcode[pkt->opcode].length. The
rxe_opcode[] array has 256 entries but is only populated for defined IB
opcodes; any other entry (for example opcode 0xff) is zero-initialized, so
length == 0 and the check degenerates to
pkt->paylen < 0 + bth_pad(pkt) + RXE_ICRC_SIZE
which does not constrain pkt->paylen enough. rxe_icrc_hdr() then computes
rxe_opcode[pkt->opcode].length - RXE_BTH_BYTES
which underflows when length == 0 and passes a huge value to rxe_crc32(),
causing an out-of-bounds read of the skb payload.
Reproduced on v7.0-rc7 with that fix applied, QEMU/KVM with
CONFIG_RDMA_RXE=y and CONFIG_KASAN=y, after
rdma link add rxe0 type rxe netdev eth0
A single 48-byte UDP packet to port 4791 with BTH opcode=0xff and
QPN=IB_MULTICAST_QPN triggers:
BUG: KASAN: slab-out-of-bounds in crc32_le+0x115/0x170
Read of size 1 at addr ...
The buggy address is located 0 bytes to the right of
allocated 704-byte region
Call Trace:
crc32_le+0x115/0x170
rxe_icrc_hdr.isra.0+0x226/0x300
rxe_icrc_check+0x13f/0x3a0
rxe_rcv+0x6e1/0x16e0
rxe_udp_encap_recv+0x20a/0x320
udp_queue_rcv_one_skb+0x7ed/0x12c0
Subsequent packets with the same shape fault on unmapped memory and panic
the kernel. The trigger requires only module load and "rdma link add"; no
QP, no connection, and no authentication.
Fix this by rejecting packets whose opcode has no rxe_opcode[] entry,
detected via the zero mask or zero length, before any length arithmetic
runs.
Cc: stable@vger.kernel.org Fixes: 8700e3e7c485 ("Soft RoCE driver") Link: https://patch.msgid.link/r/20260414111555.3386793-1-michael.bommarito@gmail.com Assisted-by: Claude:claude-opus-4-6 Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com> Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
atomic_write_reply() at drivers/infiniband/sw/rxe/rxe_resp.c
unconditionally dereferences 8 bytes at payload_addr(pkt):
value = *(u64 *)payload_addr(pkt);
check_rkey() previously accepted an ATOMIC_WRITE request with pktlen ==
resid == 0 because the length validation only compared pktlen against
resid. A remote initiator that sets the RETH length to 0 therefore reaches
atomic_write_reply() with a zero-byte logical payload, and the responder
reads sizeof(u64) bytes from past the logical end of the packet into
skb->head tailroom, then writes those 8 bytes into the attacker's MR via
rxe_mr_do_atomic_write(). That is a remote disclosure of 4 bytes of kernel
tailroom per probe (the other 4 bytes are the packet's own trailing ICRC).
IBA oA19-28 defines ATOMIC_WRITE as exactly 8 bytes. Anything else is
protocol-invalid. Hoist a strict length check into check_rkey() so the
responder never reaches the unchecked dereference, and keep the existing
WRITE-family length logic for the normal RDMA WRITE path.
Reproduced on mainline with an unmodified rxe driver: a sustained
zero-length ATOMIC_WRITE probe repeatedly leaks adjacent skb head-buffer
bytes into the attacker's MR, including recognisable kernel strings and
partial kernel-direct-map pointer words. With this patch applied the
responder rejects the PDU and the MR stays all-zero.
Cc: stable@vger.kernel.org Fixes: 034e285f8b99 ("RDMA/rxe: Make responder support atomic write on RC service") Link: https://patch.msgid.link/r/20260418162141.3610201-1-michael.bommarito@gmail.com Assisted-by: Claude:claude-opus-4-7 Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com> Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sashiko points out that pd->uctx isn't initialized until late in the
function so all these error flow references are NULL and will crash. Use
the uctx that isn't NULL.
mlx5_ib_dev_res_srq_init() allocates two SRQs, s0 and s1. When
ib_create_srq() fails for s1, the error branch destroys s0 but falls
through and unconditionally assigns the freed s0 and the ERR_PTR s1 to
devr->s0 and devr->s1.
This leads to several problems: the lock-free fast path checks
"if (devr->s1) return 0;" and treats the ERR_PTR as already initialised;
users in mlx5_ib_create_qp() dereference the freed SRQ or ERR_PTR via
to_msrq(devr->s0)->msrq.srqn; and mlx5_ib_dev_res_cleanup() dereferences
the ERR_PTR and double-frees s0 on teardown.
Fix by adding the same `goto unlock` in the s1 failure path.
Cc: stable@vger.kernel.org Fixes: 5895e70f2e6e ("IB/mlx5: Allocate resources just before first QP/SRQ is created") Link: https://patch.msgid.link/r/SYBPR01MB7881E1E0970268BD69C0BA75AF2B2@SYBPR01MB7881.ausprd01.prod.outlook.com Reported-by: Yuhao Jiang <danisjiang@gmail.com> Signed-off-by: Junrui Luo <moonafterrain@outlook.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sashiko points out the radix_tree itself is RCU safe, but nothing ever
frees the mlx4_srq struct with RCU, and it isn't even accessed within the
RCU critical section. It also will crash if an event is delivered before
the srq object is finished initializing.
Use the spinlock since it isn't easy to make RCU work, use
refcount_inc_not_zero() to protect against partially initialized objects,
and order the refcount_set() to be after the srq is fully initialized.
Sashiko points out that rx_hash_key_len comes from a uAPI structure and is
blindly passed to memcpy, allowing the userspace to trash kernel
memory. Bounds check it so the memcpy cannot overflow.
Sashiko points out that the user can specify WQs sharing the same CQ as a
part of the uAPI and this will trigger the WARN_ON() then go on to corrupt
the kernel.
node_desc[64] in struct ib_device is not guaranteed to be NUL-
terminated. The core IB sysfs handler uses "%.64s" for exactly this
reason (drivers/infiniband/core/sysfs.c:1307), since node_desc_store()
performs a raw memcpy of up to IB_DEVICE_NODE_DESC_MAX bytes with no NUL
termination:
If exactly 64 bytes are written via the node_desc sysfs file, the array
contains no NUL byte. The ionic hca_type_show() handler uses unbounded
"%s" and will read past the end of node_desc into adjacent fields of
struct ib_device until it encounters a NUL.
ionic supports IB_DEVICE_MODIFY_NODE_DESC, so this is triggerable by
userspace.
Match the core handler and bound the format specifier.
Cc: stable@vger.kernel.org Fixes: 2075bbe8ef03 ("RDMA/ionic: Register device ops for miscellaneous functionality") Link: https://patch.msgid.link/r/CALynFi7NAbhDCt1tdaDbf6TnLvAqbaHa6-Wqf6OkzREbA_PAfg@mail.gmail.com Signed-off-by: Kai Aizen <kai.aizen.dev@gmail.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Panther cove µarch starts to support auto counter reload (ACR), but the
static_call intel_pmu_enable_acr_event() is not updated for the Panther
Cove µarch used by DMR. It leads to the auto counter reload is not
really enabled on DMR.
Update static_call intel_pmu_enable_acr_event() in intel_pmu_init_pnc().
Fixes: d345b6bb8860 ("perf/x86/intel: Add core PMU support for DMR") Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20260430002558.712334-5-dapeng1.mi@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On platforms with Auto Counter Reload (ACR) support, such as NVL, a
"NMI received for unknown reason 30" warning is observed when running
multiple events in a group with ACR enabled:
$ perf record -e '{instructions/period=20000,acr_mask=0x2/u,\
cycles/period=40000,acr_mask=0x3/u}' ./test
The warning occurs because the Performance Monitoring Interrupt (PMI)
is enabled for the self-reloaded event (the cycles event in this case).
According to the Intel SDM, the overflow bit
(IA32_PERF_GLOBAL_STATUS.PMCn_OVF) is never set for self-reloaded events.
Since the bit is not set, the perf NMI handler cannot identify the source
of the interrupt, leading to the "unknown reason" message.
Furthermore, enabling PMI for self-reloaded events is unnecessary and
can lead to extraneous records that pollute the user's requested data.
Disable the interrupt bit for all events configured with ACR self-reload.
Fixes: ec980e4facef ("perf/x86/intel: Support auto counter reload") Reported-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20260430002558.712334-4-dapeng1.mi@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Members of an ACR group are logically linked via a bitmask of their
hardware counter indices. If some members of the group are assigned new
hardware counters during rescheduling, even events that keep their
original counter index must be updated with a new mask.
Without this, an event will continue to use a stale acr_mask that
references the old indices of its group peers. Ensure all ACR events are
reprogrammed during the scheduling path to maintain consistency across
the group.
Fixes: ec980e4facef ("perf/x86/intel: Support auto counter reload") Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20260430002558.712334-3-dapeng1.mi@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This occurs when allocating MSI-X vectors for an NVMe device. During
allocation the XIVE code creates a struct xive_irq_data and stores it
in irq_data->chip_data.
When the MSI-X irqdomain is later freed, xive_irq_free_data() is
responsible for retrieving this structure and freeing it. However,
after commit cc0cc23babc9 ("powerpc/xive: Untangle xive from child
interrupt controller drivers"), xive_irq_free_data() retrieves the
chip_data using irq_get_chip_data(), which looks up the data through
the child domain.
This is incorrect because the XIVE-specific irq data is associated with
the XIVE (parent) domain. As a result the lookup fails and the allocated
struct xive_irq_data is never freed, leading to the kmemleak report
shown above.
Fix this by retrieving the irq_data from the correct domain using
irq_domain_get_irq_data() and then accessing the chip_data via
irq_data_get_irq_chip_data().
If vmax has the default value of INT_MAX (e.g. because not specified in
DT), battery health is reported as over-voltage. This is because adding
any value to vmax (the vmax tolerance in this case) causes it to wrap
around, making it negative and smaller than the measured battery
voltage.
When aspm_calc_l12_info() programs the L1 PM Substates Control 1 register
fields Common_Mode_Restore_Time, LTR_L1.2_THRESHOLD_Value and _Scale, it
invokes pci_clear_and_set_config_dword() in an incorrect way:
For the bits to clear it selects those corresponding to the field. So far
so good. But for the bits to set it passes a full register value.
pci_clear_and_set_config_dword() performs a boolean OR operation which
sets all bits of that value, not just the ones that were just cleared.
Thus, when setting the LTR_L1.2_THRESHOLD_Value and _Scale on the child of
an ASPM link, aspm_calc_l12_info() also sets the Common_Mode_Restore_Time.
That's a spec violation: PCIe r7.0 sec 7.8.3.3 says this field is RsvdP
for Upstream Ports. On Adrià 's Pixelbook Eve, Common_Mode_Restore_Time
of the Intel 7265 "Stone Peak" wifi card is zero, yet aspm_calc_l12_info()
does not preserve the zero bits but instead programs the value calculated
for the Root Port into the wifi card.
Likewise, when setting the Common_Mode_Restore_Time on the Root Port,
aspm_calc_l12_info() also changes the LTR_L1.2_THRESHOLD_Value and _Scale
from the initial 163840 nsec to 237568 nsec (due to ORing those fields),
only to reduce it afterwards to 106496 nsec.
Amend all invocations of pci_clear_and_set_config_dword() to only set bits
which are cleared.
Finally, when setting the T_POWER_ON_Value and _Scale on the Root Port and
the wifi card, aspm_calc_l12_info() fails to preserve bits declared RsvdP
and instead overwrites them with zeroes. Replace pci_write_config_dword()
with pci_clear_and_set_config_dword() to avoid this.
When searching for the error source, the AER driver rules out devices whose
enable_cnt is zero. This was introduced in 2009 by commit 28eb27cf0839
("PCI AER: support invalid error source IDs") without providing a
rationale.
Drivers typically call pci_enable_device() on probe, hence the enable_cnt
check essentially filters out unbound devices. At the time of the commit,
drivers had to opt in to AER by calling pci_enable_pcie_error_reporting()
and so any AER-enabled device could be assumed to be bound to a driver.
The check thus made sense because it allowed skipping config space accesses
to devices which were known not to be the error source.
But since 2022, AER is universally enabled on all devices when they are
enumerated, cf. commit f26e58bf6f54 ("PCI/AER: Enable error reporting when
AER is native").
Errors may very well be reported by unbound devices, e.g. due to link
instability. By ruling them out as error source, errors reported by them
are neither logged nor cleared. When they do get bound and another error
occurs, the earlier error is reported together with the new error, which
may confuse users. Stop doing so.
Currently, pcie_clear_device_status() clears the entire PCIe Device Status
register (PCI_EXP_DEVSTA) by writing back the value read from the register,
which affects not only the error status bits but also other writable bits.
According to PCIe r7.0, sec 7.5.3.5, this register contains:
- RW1C error status bits (CED, NFED, FED, URD at bits 0-3): These are the
four error status bits that need to be cleared.
- Read-only bits (AUXPD at bit 4, TRPND at bit 5): Writing to these has
no effect.
- Emergency Power Reduction Detected (bit 6): A RW1C non-error bit
introduced in PCIe r5.0 (2019). This is currently the only writable
non-error bit in the Device Status register. Unconditionally clearing
this bit can interfere with other software components that rely on this
power management indication.
- Reserved bits (RsvdZ): These bits are required to be written as zero.
Writing 1s to them (as the current implementation may do) violates the
specification.
To prevent unintended side effects, modify pcie_clear_device_status() to
only write 1s to the four error status bits (CED, NFED, FED, URD), leaving
the Emergency Power Reduction Detected bit and reserved bits unaffected.
Bernd reports passthrough failure of a Digital Devices Cine S2 V6 DVB
adapter plugged into an ASRock X570S PG Riptide board with BIOS version
P5.41 (09/07/2023):
ddbridge 0000:05:00.0: detected Digital Devices Cine S2 V6 DVB adapter
ddbridge 0000:05:00.0: cannot read registers
ddbridge 0000:05:00.0: fail
BIOS assigns an incorrect BAR to the DVB adapter which doesn't fit into the
upstream bridge window. The kernel corrects the BAR assignment:
pci 0000:07:00.0: BAR 0 [mem 0xfffffffffc500000-0xfffffffffc50ffff 64bit]: can't claim; no compatible bridge window
pci 0000:07:00.0: BAR 0 [mem 0xfc500000-0xfc50ffff 64bit]: assigned
Correction of the BAR assignment happens in an x86-specific fs_initcall,
pcibios_assign_resources(), after device enumeration in a subsys_initcall.
This order was introduced at the behest of Linus in 2004:
No other architecture performs such a late BAR correction.
Bernd bisected the issue to commit a2f1e22390ac ("PCI/ERR: Ensure error
recoverability at all times"), but it only occurs in the absence of commit 4d4c10f763d7 ("PCI: Explicitly put devices into D0 when initializing").
This combination exists in stable kernel v6.12.70, but not in mainline,
hence Bernd cannot reproduce the issue with mainline.
Since a2f1e22390ac, config space is saved on enumeration, prior to BAR
correction. Upon passthrough, the corrected BAR is overwritten with the
incorrect saved value by:
But only if the device's current_state is PCI_UNKNOWN, as it was prior to
commit 4d4c10f763d7. Since the commit, it is PCI_D0, which changes the
behavior of vfio_pci_set_power_state() to no longer restore the state
without saving it first.
Alexandre is reporting the same issue as Bernd, but in his case, mainline
is affected as well. The difference is that on Alexandre's system, the
host kernel binds a driver to the device which is unbound prior to
passthrough, whereas on Bernd's system no driver gets bound by the host
kernel.
Unbinding sets current_state to PCI_UNKNOWN in pci_device_remove(), so when
vfio-pci is subsequently bound to the device, pci_restore_state() is once
again called without invoking pci_save_state() first.
To robustly fix the issue, always update saved_config_space upon resource
assignment.
Reported-by: Bernd Schumacher <bernd@bschu.de> Closes: https://lore.kernel.org/r/acfZrlP0Ua_5D3U4@eldamar.lan/ Reported-by: Alexandre N. <an.tech@mailo.com> Closes: https://lore.kernel.org/r/dd3c3358-de0f-4a56-9c81-04aceaab4058@mailo.com/ Fixes: a2f1e22390ac ("PCI/ERR: Ensure error recoverability at all times") Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Tested-by: Bernd Schumacher <bernd@bschu.de> Tested-by: Alexandre N. <an.tech@mailo.com> Cc: stable@vger.kernel.org # v6.12+ Link: https://patch.msgid.link/febc3f354e0c1f5a9f5b3ee9ffddaa44caccf651.1776268054.git.lukas@wunner.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
damon_sysfs_quot_goal->path can be read and written by users, via DAMON
sysfs 'path' file. It can also be indirectly read, for the parameters
{on,off}line committing to DAMON. The reads for parameters committing are
protected by damon_sysfs_lock to avoid the sysfs files being destroyed
while any of the parameters are being read. But the user-driven direct
reads and writes are not protected by any lock, while the write is
deallocating the path-pointing buffer. As a result, the readers could
read the already freed buffer (user-after-free). Note that the user-reads
don't race when the same open file is used by the writer, due to kernfs's
open file locking. Nonetheless, doing the reads and writes with separate
open files would be common. Fix it by protecting both the user-direct
reads and writes with damon_sysfs_lock.
Patch series "mm/damon/sysfs-schemes: fix use-after-free for [memcg_]path".
Reads of 'memcg_path' and 'path' files in DAMON sysfs interface could race
with their writes, results in use-after-free. Fix those.
This patch (of 2):
damon_sysfs_scheme_filter->mmecg_path can be read and written by users,
via DAMON sysfs memcg_path file. It can also be indirectly read, for the
parameters {on,off}line committing to DAMON. The reads for parameters
committing are protected by damon_sysfs_lock to avoid the sysfs files
being destroyed while any of the parameters are being read. But the
user-driven direct reads and writes are not protected by any lock, while
the write is deallocating the memcg_path-pointing buffer. As a result,
the readers could read the already freed buffer (user-after-free). Note
that the user-reads don't race when the same open file is used by the
writer, due to kernfs's open file locking. Nonetheless, doing the reads
and writes with separate open files would be common. Fix it by protecting
both the user-direct reads and writes with damon_sysfs_lock.
Link: https://lore.kernel.org/20260423150253.111520-1-sj@kernel.org Link: https://lore.kernel.org/20260423150253.111520-2-sj@kernel.org Fixes: 4f489fe6afb3 ("mm/damon/sysfs-schemes: free old damon_sysfs_scheme_filter->memcg_path on write") Co-developed-by: Junxi Qian <qjx1298677004@gmail.com> Signed-off-by: Junxi Qian <qjx1298677004@gmail.com> Signed-off-by: SeongJae Park <sj@kernel.org> Cc: <stable@vger.kernel.org> # 6.16.x Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DAMON_STAT updates 'enabled' parameter value, which represents the running
status of its kdamond, when the user explicitly requests start/stop of the
kdamond. The kdamond can, however, be stopped even if the user explicitly
requested the stop, if ctx->regions_score_histogram allocation failure at
beginning of the execution of the kdamond. Hence, if the kdamond is
stopped by the allocation failure, the value of the parameter can be
stale.
Users could show the stale value and be confused. The problem will only
rarely happen in real and common setups because the allocation is arguably
too small to fail. Also, unlike the similar bugs that are now fixed in
DAMON_RECLAIM and DAMON_LRU_SORT, kdamond can be restarted in this case,
because DAMON_STAT force-updates the enabled parameter value for user
inputs. The bug is a bug, though.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
Patch series "mm/damon/modules: detect and use fresh status", v3.
DAMON modules including DAMON_RECLAIM, DAMON_LRU_SORT and DAMON_STAT
commonly expose the kdamond running status via their parameters. Under
certain scenarios including wrong user inputs and memory allocation
failures, those parameter values can be stale. It can confuse users. For
DAMON_RECLAIM and DAMON_LRU_SORT, it even makes the kdamond unable to be
restarted before the system reboot.
The problem comes from the fact that there are multiple events for the
status changes and it is difficult to follow up all the scenarios. Fix
the issue by detecting and using the status on demand, instead of using a
cached status that is difficult to be updated.
Patches 1-3 fix the bugs in DAMON_RECLAIM, DAMON_LRU_SORT and DAMON_STAT
in the order.
This patch (of 3):
DAMON_RECLAIM updates 'enabled' and 'kdamond_pid' parameter values, which
represents the running status of its kdamond, when the user explicitly
requests start/stop of the kdamond. The kdamond can, however, be stopped
in events other than the explicit user request in the following three
events.
1. ctx->regions_score_histogram allocation failure at beginning of the
execution,
2. damon_commit_ctx() failure due to invalid user input, and
3. damon_commit_ctx() failure due to its internal allocation failures.
Hence, if the kdamond is stopped by the above three events, the values of
the status parameters can be stale. Users could show the stale values and
be confused. This is already bad, but the real consequence is worse.
DAMON_RECLAIM avoids unnecessary damon_start() and damon_stop() calls
based on the 'enabled' parameter value. And the update of 'enabled'
parameter value depends on the damon_start() and damon_stop() call
results. Hence, once the kdamond has stopped by the unintentional events,
the user cannot restart the kdamond before the system reboot. For
example, the issue can be reproduced via below steps.
The problem will only rarely happen in real and common setups for the
following reasons. The allocation failures are unlikely in such setups
since those allocations are arguably too small to fail. Also sane users
on real production environments may not commit wrong input parameters.
But once it happens, the consequence is quite bad. And the bug is a bug.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
Link: https://lore.kernel.org/20260419161003.79176-1-sj@kernel.org Link: https://lore.kernel.org/20260419161003.79176-2-sj@kernel.org Fixes: e035c280f6df ("mm/damon/reclaim: support online inputs update") Co-developed-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: SeongJae Park <sj@kernel.org> Cc: <stable@vger.kernel.org> # 5.19.x Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DAMON_LRU_SORT updates 'enabled' and 'kdamond_pid' parameter values, which
represents the running status of its kdamond, when the user explicitly
requests start/stop of the kdamond. The kdamond can, however, be stopped
in events other than the explicit user request in the following three
events.
1. ctx->regions_score_histogram allocation failure at beginning of the
execution,
2. damon_commit_ctx() failure due to invalid user input, and
3. damon_commit_ctx() failure due to its internal allocation failures.
Hence, if the kdamond is stopped by the above three events, the values of
the status parameters can be stale. Users could show the stale values and
be confused. This is already bad, but the real consequence is worse.
DAMON_LRU_SORT avoids unnecessary damon_start() and damon_stop() calls
based on the 'enabled' parameter value. And the update of 'enabled'
parameter value depends on the damon_start() and damon_stop() call
results. Hence, once the kdamond has stopped by the unintentional events,
the user cannot restart the kdamond before the system reboot. For
example, the issue can be reproduced via below steps.
The problem will only rarely happen in real and common setups for the
following reasons. The allocation failures are unlikely in such setups
since those allocations are arguably too small to fail. Also sane users
on real production environments may not commit wrong input parameters.
But once it happens, the consequence is quite bad. And the bug is a bug.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
Link: https://lore.kernel.org/20260419161003.79176-3-sj@kernel.org Fixes: 40e983cca927 ("mm/damon: introduce DAMON-based LRU-lists Sorting") Co-developed-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: SeongJae Park <sj@kernel.org> Cc: <stable@vger.kernel.org> # 6.0.x Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fall back to apic_find_highest_vector() when PID.ON is set but PIR
turns out to be empty, to correctly report the highest pending interrupt
from the existing IRR.
In a nested VM stress test, the following WARNING fires in
vmx_check_nested_events() when kvm_cpu_has_interrupt() reports a pending
interrupt but the subsequent kvm_apic_has_interrupt() (which invokes
vmx_sync_pir_to_irr() again) returns -1:
The root cause is a race between vmx_sync_pir_to_irr() on the target vCPU
and __vmx_deliver_posted_interrupt() on a sender vCPU. The sender
performs two individually-atomic operations that are not a single
transaction:
1. pi_test_and_set_pir(vector) -- sets the PIR bit
2. pi_test_and_set_on() -- sets PID.ON
The following interleaving triggers the bug:
Sender vCPU (IPI): Target vCPU (1st sync_pir_to_irr):
B1: set PIR[vector]
A1: pi_clear_on()
A2: pi_harvest_pir() -> sees B1 bit
A3: xchg() -> consumes bit, PIR=0
(1st sync returns correct max_irr)
B2: set PID.ON = 1
Target vCPU (2nd sync_pir_to_irr):
C1: pi_test_on() -> TRUE (from B2)
C2: pi_clear_on() -> ON=0
C3: pi_harvest_pir() -> PIR empty
C4: *max_irr = -1, early return
IRR NOT SCANNED
The interrupt is not lost (it resides in the IRR from the first sync and
is recovered on the next vcpu_enter_guest() iteration), but the incorrect
max_irr causes a spurious WARNING and a wasted L2 VM-Enter/VM-Exit cycle.
Checking is_guest_mode(vcpu) is incorrect, because translate_nested_gpa()
is only valid if an L2 guest is running *with nested EPT/NPT enabled*.
Instead use the same condition as translate_nested_gpa() itself.
Cc: stable@vger.kernel.org Reviewed-by: Sean Christopherson <seanjc@google.com> Fixes: aee738236dca ("KVM: x86: Prepare kvm_hv_flush_tlb() to handle L2's GPAs", 2022-11-18) Link: https://patch.msgid.link/20260503200905.106077-1-pbonzini@redhat.com/ Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
parse_sec_desc(), build_sec_desc(), and the chown path in
id_mode_to_cifs_acl() all add the server-supplied dacloffset to pntsd
before proving a DACL header fits inside the returned security
descriptor.
On 32-bit builds a malicious server can return dacloffset near
U32_MAX, wrap the derived DACL pointer below end_of_acl, and then slip
past the later pointer-based bounds checks. build_sec_desc() and
id_mode_to_cifs_acl() can then dereference DACL fields from the wrapped
pointer in the chmod/chown rewrite paths.
Validate dacloffset numerically before building any DACL pointer and
reuse the same helper at the three DACL entry points.
Fixes: bc3e9dd9d104 ("cifs: Change SIDs in ACEs while transferring file ownership.") Cc: stable@vger.kernel.org Assisted-by: Claude:claude-opus-4-6 Signed-off-by: Michael Bommarito <michael.bommarito@gmail.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 62e7dd0a39c2d ("smb: common: change the data type of num_aces
to le16") split struct smb_acl's __le32 num_aces field into __le16
num_aces and __le16 reserved. The reserved field corresponds to Sbz2
in the MS-DTYP ACL wire format, which must be zero [1].
When building an ACL descriptor in build_sec_desc(), we are using a
kmalloc()'ed descriptor buffer and writing the fields explicitly using
le16() writes now. This never writes to the 2 byte reserved field,
leaving it as uninitialized heap data.
When the reserved field happens to contain non-zero slab garbage,
Samba rejects the security descriptor with "ndr_pull_security_descriptor
failed: Range Error", causing chmod to fail with EINVAL.
Change kmalloc() to kzalloc() to ensure the entire buffer is
zero-initialized.
Fixes: 62e7dd0a39c2d ("smb: common: change the data type of num_aces to le16") Cc: stable@vger.kernel.org Signed-off-by: Bjoern Doebel <doebel@amazon.de> Assisted-by: Kiro:claude-opus-4.6
[1] https://learn.microsoft.com/en-us/openspecs/windows_protocols/ms-dtyp/20233ed8-a6c6-4097-aafa-dd545ed24428 Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / thirdparty / kernel / stable.git / log
